Affects the opennlp-uima package, in particular the opennlp/uima/postag/POSTaggerTrainer.java class.
This AE is expected to parse token annotations and to build two data structures. The first one is an array of the token coveredTexts and the second an array of associated tags (the tags are specified by a feature structure path set in parameter).
In practice, the tag value of the current token is wrongly added to the token array.
This can be easily solved by changing the name of the data structure: from `tokens` to `tags` at line 200.